hierarchical
A hierarchical delegation framework where a lead agent directs worker agents to complete tasks.
Install
mkdir -p .claude/skills/hierarchical && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/14154" && unzip -o skill.zip -d .claude/skills/hierarchical && rm skill.zipInstalls to .claude/skills/hierarchical
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Boss delegates tasks to workers, reviews their output, and approves or requests changes.Key capabilities
- →Delegate specific tasks to worker agents
- →Review worker agent submissions
- →Approve or reject work with feedback
- →Iterate on revisions with worker agents
- →Signal completion of all approved work
How it works
A boss agent delegates tasks to worker agents, reviews their output, and provides feedback for revisions until all work is approved, following a clear communication protocol.
Inputs & outputs
When to use hierarchical
- →Coordinating multi-agent coding tasks
- →Managing task delegation
- →Implementing iterative review cycles
- →Structuring agent project work
About this skill
Hierarchical
Purpose
A structured delegation pattern where a boss agent assigns work to worker agents, reviews their output, and makes final approval decisions. Clear chain of command.
Use mesh-comms-core first if direct worker notification has not already been proven. comms.md is the ledger, not the wake-up path.
Default roster: use Claude Code and Codex peers. Do not assign Gemini CLI, Antigravity CLI, or Antigravity Desktop as boss or worker unless the human prompt explicitly names that peer.
Roles
| Role | Description |
|---|---|
BOSS | Delegates tasks, reviews work, approves or rejects |
WORKER | Receives assignment, executes, submits for review |
One BOSS, one or more WORKERs.
Protocol
1. DELEGATION: BOSS assigns specific tasks to each WORKER
2. EXECUTION: WORKERs complete their assigned tasks
3. SUBMISSION: WORKERs submit work with [SUBMIT]
4. REVIEW: BOSS reviews each submission
5. DECISION: BOSS signals [APPROVE] or [REVISE: feedback]
6. ITERATE: If revisions needed, WORKERs address and resubmit
7. COMPLETE: When all work approved, BOSS signals [DONE]
Delegation Format
[DELEGATION - BOSS]
@WORKER_1: [Task description]
@WORKER_2: [Task description]
DEADLINE: [Turn count or "when ready"]
ACCEPTANCE CRITERIA: [What "done" looks like]
Submission Format
[SUBMIT - WORKER]
TASK: [What was assigned]
DELIVERABLE: [The work product]
NOTES: [Any caveats or questions]
Review Format
[REVIEW - BOSS]
@WORKER: [APPROVE] or [REVISE]
FEEDBACK: [What's good, what needs work]
Inputs
All inputs come from the prompt.
Optional:
output: File path for final deliverablework: File containing the overall goalmax_turns: Override default (30)
Prompt Format Examples
Minimal:
Use hierarchical to build a landing page.
CC is BOSS. CX and CC-peer-02 are WORKERs.
With specific tasks:
Use hierarchical for code review workflow.
BOSS=CC (Tech Lead).
WORKER=CX (implements feature).
WORKER=CC-peer-02 (writes tests).
Output: projects/feature/implementation.md
Safety
- max_turns: 30 (default)
- Stuck reviews: If BOSS doesn't respond, WORKERs may escalate with [ESCALATE]
- Status logged to: direct notification plus
comms.md
Example Session
[DELEGATION - CC (BOSS)]
Project: Create a press release for our new AI product.
@CX: Write the headline and first paragraph (hook)
@CC-peer-02: Write the product details and features section
ACCEPTANCE CRITERIA:
- Compelling headline (under 10 words)
- Clear value proposition
- Three key features highlighted
[SUBMIT - CX]
TASK: Headline and hook
DELIVERABLE:
# AI Agents That Actually Collaborate
Today we announce Interlateral, the first multi-agent framework
where AI systems work together like a real team.
NOTES: Went for impact over cleverness.
[SUBMIT - CC-peer-02]
TASK: Product details and features
DELIVERABLE:
## Key Features
1. **Natural Collaboration** - Agents communicate in plain language
2. **Design Patterns** - Pre-built patterns for common workflows
3. **Observable** - Full visibility into agent interactions
NOTES: Kept it to three as requested.
[REVIEW - CC (BOSS)]
@CX: [APPROVE] - Headline is strong. Good hook.
@CC-peer-02: [REVISE] - Add one more feature about safety/reliability.
[SUBMIT - CC-peer-02]
TASK: Revised features
DELIVERABLE:
## Key Features
1. **Natural Collaboration** - Agents communicate in plain language
2. **Design Patterns** - Pre-built patterns for common workflows
3. **Observable** - Full visibility into agent interactions
4. **Fail-Safe** - Built-in safeguards prevent runaway processes
NOTES: Added safety feature as requested.
[REVIEW - CC (BOSS)]
@CC-peer-02: [APPROVE] - Perfect.
[DONE] - All work approved. Compiling final document.
Final Status Format
SKILL: hierarchical
STATUS: DONE
TURNS: 6
OUTPUT: projects/press/release.md
PARTICIPANTS: CC (BOSS), CX (WORKER), CC-peer-02 (WORKER)
SUBMISSIONS: 3
APPROVALS: 3
REVISIONS: 1
When not to use it
- →When direct worker notification has not already been proven
Limitations
- →Two to four explicitly selected agents.
- →Default to Claude Code + Codex peers; Gemini and Antigravity are opt-in only when the human explicitly selects them.
- →max_turns: 30 (default)
How it compares
This skill establishes a structured, hierarchical delegation and review process for agents, ensuring a clear chain of command and iterative feedback, unlike unstructured collaboration.
Compared to similar skills
hierarchical side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| hierarchical (this skill) | 0 | 2mo | No flags | Intermediate |
| trello | 41 | 2mo | Review | Beginner |
| executing-plans | 6 | 3mo | No flags | Intermediate |
| github-project-management | 4 | 6mo | Review | Advanced |
Try saying
Example prompts that trigger this skill in your AI assistant.
More by dazzaji
View all by dazzaji →You might also like
trello
openclaw
Manage Trello boards, lists, and cards via the Trello REST API.
executing-plans
obra
Use when you have a written implementation plan to execute in a separate session with review checkpoints
github-project-management
ruvnet
Comprehensive GitHub project management with swarm-coordinated issue tracking, project board automation, and sprint planning
project-clickup
incidentfox
ClickUp project management integration for incident tracking and task management
coo-advisor
alirezarezvani
Operations leadership for scaling companies. Process design, OKR execution, operational cadence, and scaling playbooks. Use when designing operations, setting up OKRs, building processes, scaling teams, analyzing bottlenecks, planning operational cadence, or when user mentions COO, operations, process improvement, OKRs, scaling, operational efficiency, or execution.
tlc-spec-driven
tech-leads-club
Project and feature planning with 4 phases - Specify, Design, Tasks, Implement+Validate. Creates atomic tasks with verification criteria and maintains persistent memory across sessions. Stack-agnostic. Use when: (1) Starting new projects (initialize vision, goals, roadmap), (2) Working with existing codebases (map stack, architecture, conventions), (3) Planning features (requirements, design, task breakdown), (4) Implementing with verification, (5) Tracking decisions/blockers across sessions, (6) Pausing/resuming work. Triggers on "initialize project", "map codebase", "specify feature", "design", "tasks", "implement", "pause work", "resume work".